{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:F7D3V4QPWP7OWH3FGCYFNPWK2X","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"9d11ab0048ddf714e0550a92ef28b9b600db89111931cbb6ba79e1f77f1fc3d2","cross_cats_sorted":["cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-02-19T17:01:08Z","title_canon_sha256":"2c3764cb9431ee8c37b47e285be808e3e1a10e56063d48b7f700fdbdbef3690f"},"schema_version":"1.0","source":{"id":"2602.17547","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2602.17547","created_at":"2026-06-10T12:16:10Z"},{"alias_kind":"arxiv_version","alias_value":"2602.17547v3","created_at":"2026-06-10T12:16:10Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.17547","created_at":"2026-06-10T12:16:10Z"},{"alias_kind":"pith_short_12","alias_value":"F7D3V4QPWP7O","created_at":"2026-06-10T12:16:10Z"},{"alias_kind":"pith_short_16","alias_value":"F7D3V4QPWP7OWH3F","created_at":"2026-06-10T12:16:10Z"},{"alias_kind":"pith_short_8","alias_value":"F7D3V4QP","created_at":"2026-06-10T12:16:10Z"}],"graph_snapshots":[{"event_id":"sha256:3af6fb5885fc95cf0bf052d03bcebb9ebfd03b0ad63d16f5747aee80812da326","target":"graph","created_at":"2026-06-10T12:16:10Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"our proposed KLong (106B) surpasses Kimi K2 Thinking (1T) by 11.28% on PaperBench, and the performance improvement generalizes to other coding benchmarks like SWE-bench Verified and MLE-bench."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That trajectories distilled from Claude 4.5 Sonnet via the Research-Factory pipeline are high-quality and that trajectory-splitting SFT plus progressive RL stages preserve sufficient context and capability for true long-horizon generalization without hidden failure modes."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"KLong trains a 106B LLM agent using trajectory-splitting SFT and progressive RL to outperform a 1T model by 11.28% on PaperBench with generalization to coding benchmarks."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"KLong shows that trajectory-splitting SFT followed by progressive RL lets a 106B agent outperform a 1T model on extremely long-horizon tasks."}],"snapshot_sha256":"672e21a8ba139cfd1517a415d54751d0de2d8bc6c4efe0145f7f084dda03212d"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2602.17547/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"This paper introduces KLong, an open-source LLM agent trained to solve extremely long-horizon tasks. The principle is to first cold-start the model via trajectory-splitting SFT, then scale it via progressive RL training. Specifically, we first activate basic agentic abilities of a base model with a comprehensive SFT recipe. Then, we introduce Research-Factory, an automated pipeline that generates high-quality training data by collecting research papers and constructing evaluation rubrics. Using this pipeline, we build thousands of long-horizon trajectories distilled from Claude 4.5 Sonnet (Thi","authors_text":"Bryan Hooi, Flood Sung, Jiaheng Zhang, Xinlong Yang, Yanhao Li, Yibo Miao, Yingwei Ma, Yuchong Xie, Yue Liu, Zhiyuan Hu","cross_cats":["cs.CL"],"headline":"KLong shows that trajectory-splitting SFT followed by progressive RL lets a 106B agent outperform a 1T model on extremely long-horizon tasks.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-02-19T17:01:08Z","title":"KLong: Training LLM Agent for Extremely Long-horizon Tasks"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.17547","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-15T20:45:41.123186Z","id":"31e98931-e61b-4679-8934-ce8d62549855","model_set":{"reader":"grok-4.3"},"one_line_summary":"KLong trains a 106B LLM agent using trajectory-splitting SFT and progressive RL to outperform a 1T model by 11.28% on PaperBench with generalization to coding benchmarks.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"KLong shows that trajectory-splitting SFT followed by progressive RL lets a 106B agent outperform a 1T model on extremely long-horizon tasks.","strongest_claim":"our proposed KLong (106B) surpasses Kimi K2 Thinking (1T) by 11.28% on PaperBench, and the performance improvement generalizes to other coding benchmarks like SWE-bench Verified and MLE-bench.","weakest_assumption":"That trajectories distilled from Claude 4.5 Sonnet via the Research-Factory pipeline are high-quality and that trajectory-splitting SFT plus progressive RL stages preserve sufficient context and capability for true long-horizon generalization without hidden failure modes."}},"verdict_id":"31e98931-e61b-4679-8934-ce8d62549855"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:6b780a2c22b45a44d5af5fc0e8ae70fbcb2e7b420a259c9a2e7002a26b8996ca","target":"record","created_at":"2026-06-10T12:16:10Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"9d11ab0048ddf714e0550a92ef28b9b600db89111931cbb6ba79e1f77f1fc3d2","cross_cats_sorted":["cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-02-19T17:01:08Z","title_canon_sha256":"2c3764cb9431ee8c37b47e285be808e3e1a10e56063d48b7f700fdbdbef3690f"},"schema_version":"1.0","source":{"id":"2602.17547","kind":"arxiv","version":3}},"canonical_sha256":"2fc7baf20fb3feeb1f6530b056becad5c649aa674869b745dd8e0f955258dbef","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"2fc7baf20fb3feeb1f6530b056becad5c649aa674869b745dd8e0f955258dbef","first_computed_at":"2026-06-10T12:16:10.602892Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-10T12:16:10.602892Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"gsSP9hxGmSu5MY2mxrxmgy4trcu1WQ1vrz/HGudaQppHf3NbP9TRb8tofZOczgH81DqpeTMAPtnpPnm+4bzbDQ==","signature_status":"signed_v1","signed_at":"2026-06-10T12:16:10.605544Z","signed_message":"canonical_sha256_bytes"},"source_id":"2602.17547","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:6b780a2c22b45a44d5af5fc0e8ae70fbcb2e7b420a259c9a2e7002a26b8996ca","sha256:3af6fb5885fc95cf0bf052d03bcebb9ebfd03b0ad63d16f5747aee80812da326"],"state_sha256":"61711b5e4318c1f4909afd3c3a49ed94d57559f77be5d6f4af14b4c25ad230ba"}